Skip to main content
v2 draft outline

This page is an outline only. It lists what the page will eventually cover, not final content. The current v1 course lives under Program and is unchanged.

Foundation Forge v2: Overview and Schedule

What this section is​

  • Draft outlines for the v2 course, staged beside v1
  • One page per module; each page is a list of what it will cover
  • Target ship date: Nov 30

The arc​

  • Recognize good, then fix bad, then build your own
  • Module 1: operate a known-good factory
  • Modules 2 and 3: convert a bad repo to a good one
  • Module 4: build a software factory from scratch for your own project
  • The factory is the point; bug fixes and small features are just work passing through it

Audience​

  • Nontechnical learners are in scope
  • The goal is the instinct to run a quality check before shipping, not tool mastery
  • The bar scales: nontechnical learners show the instinct and the gaps an agent found; technical learners show the full hallmarks (see Module 4)

Schedule​

4 weeks, 20 business days, 5 days per module.

WeekModuleDays
Pre-startIntroduction (unnumbered)n/a
Week 1Module 1: Operate a Good Factory5
Week 2Module 2: Diagnose the Bad Repo5
Week 3Module 3: Make It Good5
Week 4Module 4: Capstone, Build Your Factory5
  • Modules 2 and 3 are one conversion split in two: diagnose and plan, then close the gaps and prove the gates work

Practice repos​

Locked decisions that still apply​

  • Instruction file: AGENTS.md only, no CLAUDE.md
  • Coverage threshold: 75%
  • Gates in the good repo: Semgrep (SAST) and OSV-Scanner at pre-commit; Trivy (SCA) at the PR gate
  • OSV-Scanner placement: pre-commit on the good app; in the Capstone the learner chooses where it runs
  • Dependency automation: Dependabot
  • Deploy: auto-deploy to Vercel on merge, smoke test as the final gate; production on Vercel is behind Vercel login
  • Week 1 activity: learners fork the good repo, make a change, and get a green pull request in their fork
  • Platform: the company is Mac-only, so setup and exercises are written for Mac
  • Specs are context, not the spine of the course

Pages in this section​

Parked from the earlier draft​

Nothing below is in the current shape. It is kept here so it is not silently lost; some of it may be reused (noted where likely).

  • Seeded issues as the core exercise: 10 seeded issues, learners pick a subset; themed slices; 18-issue candidate pool
  • By-hand reps: at least five by-hand loop reps, with a fallback cap of 4
  • Old Module 2 skills: build your own judges (fresh-context reviewers), codify your method as a skill
  • Seeded flaws on a prepared branch (main stays clean): double booking via a checkout-date off-by-one, room detail UI bug slice, first-pass traps in specs (weekend-night definition, fee-window boundary); superseded by the Chaos seeded flaws, so check for overlap before reusing
  • Seeded scanner material: known-vulnerable dependency, injection-class and hardening-class findings, dummy credential fixture; superseded by Chaos, so check for overlap before reusing
  • Old Module 3: learner-chosen AI feature (booking assistant, guest-services concierge, or room-upgrade recommender) with model key, cost limits and a safe fallback
  • Old Module 3 extras: worth-it and token-cost check, orchestration across slices, CI size alert at about 400 changed lines, different-model review, multi-component work with a notification stub
  • Old Module 4 (Gates) as its own module: learners build tiered gates, blocked-push exercise, close a merged scanner-findings issue, evaluate their own gate, learner-defined smoke-test confidence bar, learner-designed state approach on Vercel
  • Hacktory vs. factory side-by-side rebuild as a Module 1 exercise
  • Readiness audit seed: about a 60% starting pass with AGENTS.md, test command and dependency file missing; Momentum and Wayfinding left undone for a before/after audit
  • Standards exercise placement: find in Module 1, write down in Module 2, enforce one in Module 4, re-audit in Module 5
  • Agent-hook gating awareness item (Tier 2 local review)
  • Capstone extras: reserve issue (partner availability feed), deploy target doc (AWS, Azure, Vercel, the Lab and its constraints), Terraform and IaC stretch
  • Alignment tools (grill-me style questioning, brainstorming skills) as tools to try, not gospel
  • 99% AI-written measure (see open items)

Open items​

  • TODO: confirm main of the good repo carries the factory before learners fork it
  • TODO: forks have issues turned off by default; decide how learners pick up issues in their fork (see week1-activity-options.md in the Module 1 folder, internal and not published)
  • TODO: run the Module 1 cold run on a clean Mac; the Mac setup is untested on a real Mac
  • TODO(conflict): the decisions log says the Capstone is Module 5 and the intro is unnumbered; today's shape has four modules with the Capstone as Module 4. Today's shape wins; update the log and any cross-references.
  • TODO: where the intro's time comes from inside the 20 days (it has no module slot)
  • TODO: remap the standards exercise (find, write down, enforce one, re-audit) onto Modules 2 to 4
  • TODO: is the 99% AI-written measure kept as an exit test, and for which module?
  • TODO: Vercel cost guardrails; org access is granted at Forge join
  • TODO: practice app go/no-go due Nov 25
  • TODO: Modules 2 to 4 are thin until Chaos is finalized